Invalid or non-sensical setup of inputs

[HardCoreCodin]

Some materials have input setup that makes no sense (i.e: in masonJarGlass/mtlxopen_pbr_surface):

1. Image file paths sometimes referenced from @textures/ and other times from @..texture/ and some implementations would only be able to load the latter. The default USD resolver would anchor relative paths to the directory of the file that references them, as opposed to relative to the stage root, which in the former case would have it look for the textures under /materials/textures which does not exist.
2. Color space being set srgb_tx (which does not exist in MaterialX) instead of just srgb (i.e: BaseColor/inputs:file)
3. Declaration of input without any assignment or connection (i.e: color3f inputs:geometry_opacity)
4. Inputs being set within a disabled section (i.e: coat_ior and coat_roughness being set, while coat_weight is left at 0.0)
5. Inputs being set to a value and then connected to a node (i.e: float inputs:transmission_weight)
6. Inputs of invalid names used (i.e: masonJarWater/float inputs:specular_anisotropy should be specular_roughness_anisotropy)
7. Non-sensical IOR of 1 makes materials effectively pass-through (ie: masonJarWater/float inputs:coat_ior = 1)
8. Subsurface settings of radius and scale are set on Thin Walled materials (i.e: paper) which has no effect as per the spec.
9. Emission luminance inputs are fed from images in range 0-1 where luminance values should be in the thousands (i.e: meetMat)
10. Emission luminance in meetMat is coming in containing negative values, and OpenPBR is not (yet) mandating clamping to specified ranges in the spec. It would be better to add an explicit clamp node there.
11. Orange juice group is set to be invisible in the geometry usd file, but is still being rendered when coming through the USD layer composition. This is evident in Omniverse. It is only not rendering in Arnold because of some arnold-specific primVards that disable it (which should not exist in this content at all for portability). Once the arnold-specific primvars are removed, the orange juice group is also rendered in Arnold.

[matthewlow-dwa]

Thanks @HardCoreCodin for identifying and filing these issues!

As you've likely noticed, the asset is currently in a Beta state and may not be usable or imageable out of the box.

Primarily we had been blocked by OpenUSD adoption of a version of MaterialX with OpenPBR. But that has now been resolved in the most recent release of OpenUSD v0.25.05 adopting MaterialX v1.39.3, which now enables us to return to updating this asset to the last specs in preparation for an official v1.0 release of the Shader Playground as a part of the ASWF Digital Production Example Library (DPEL) over the next few months.

We'll hope to address many of the issues you've identified as a part of that work, and PRs and additional Issues are always welcome.